Customer retention and decision services

Nancy Williams had a nice little article - New Angle on Customer Retention over on the B-eye network. Nancy makes a great point about the value of time series data in managing customer retention. I would go further, though, and say that this kind of data would allow you to build predictive models that could be embedded in the systems your staff used, typically as decision services. Combining predictive models that use this data to assess, for instance, how likely a particular complaint is to be the one that causes a customer to leave (as distinct from one that won't) with rules as to how to treat this customer to try and retain them (and indeed rules on which customers to try and retain) can make a huge difference to your operations. This requires a focus on the micro decisions of "retain this specific customer and if so how"

While you could structure the data you collect, and that is a very effective technique, text analysis technology is increasingly mature. The use of text analytics to extract variables from text to complement structured data, is growing in value for exactly this reason. In particular this approach matters in areas like collections, insurance subrogation and marketing where unstructured information can be critical. One approach is to learn relationships between words and map these to multi-dimensional vectors (this is an approach used by Fair Isaac). This allows you to take text like claims descriptions, collections notes or customer support emails and generate concept scores from them. Essentially the software analyzes the text in a free text field to create a number of "fields" for the record that contain values showing the strength of alignment to particular concepts. Modelers then combined these generated fields with traditional structured data to produce predictive models.

Technorati Tags: business rules, customer decisions, Customer Service, decision service, decision service hub, micro decisions, predictive analytics, customer retention, text analytics, unstructured data